On 27/07/2022 11:31, Kalle Valo wrote:

Bryan O'Donoghue [off-list ref] writes:

quoted

+static ssize_t read_file_firmware_feature_caps(struct file *file,
+					       char __user *user_buf,
+					       size_t count, loff_t *ppos)
+{
+	struct wcn36xx *wcn = file->private_data;
+	unsigned long page = get_zeroed_page(GFP_KERNEL);
+	char *p = (char *)page;
+	int i;
+	int ret;
+
+	if (!p)
+		return -ENOMEM;
+
+	mutex_lock(&wcn->hal_mutex);
+	for (i = 0; i < MAX_FEATURE_SUPPORTED; i++) {
+		if (wcn36xx_firmware_get_feat_caps(wcn->fw_feat_caps, i)) {
+			p += sprintf(p, "%s\n",
+				     wcn36xx_firmware_get_cap_name(i));
+		}
+	}
+	mutex_unlock(&wcn->hal_mutex);
+
+	ret = simple_read_from_buffer(user_buf, count, ppos, (char *)page,
+				      (unsigned long)p - page);
+
+	free_page(page);
+	return ret;
+}

Why not use the normal use kzalloc() and kfree()? That way you would not
need a separate page variable. What's the benefit from
get_zeroed_page()?

TBH I did a copy/paste here from another driver... I forget which

Also I don't see any checks for a memory allocation failure.

its there

char *p = (char*) page;

if (!p)
     return -ENOMEM;

I can V2 this for kzalloc and kfree if you prefer though

---
bod